Because of its many advantages, big data has been extending to various domains of science, health, education, and commerce. Despite its many applications, big data sharing typically suffers from some key issues, such as user control, lack of incentives, cost, and the right of data. This paper proposes a decentralized big data sharing prototype to improve the applications and services of big data. The method makes use of Ethereum blockchain and related technologies to systematically recommend the implementation guidelines. The research provides a detailed description of the design and implementation of each sublayer of a big data system. As the method is based on blockchain technology, the key technical points are properly addressed in each of the layers. For evaluation, relevant data were collected, and functional testing was performed. A comparison was performed about the sharing frequency and blockchain consensus performance of similar platforms. The dual mining node of the proposed prototype succeeded in processing 1366 blocks and 300 messages. A comparatively satisfactory file access time in the range of 10 m to 20 s and file transmission time between 100 m and 200 s were achieved. The results obtained show that this prototype can effectively verify the feasibility of the model, the layered architecture, and the related sharing mechanism. For the functional and performance testing, practical projects were implemented and evaluated. The promising results obtained testify that the research offers a theoretical background for innovative research in the domain and specialized guidelines for practical implementation.